Unsteady Thermal ManagementSimulations for a Passenger Vehicle using1D and 3D ToolsMaster’s thesis in Automotive Engineering

SAKET KUMARDANIEL PASCUAL

Department of Applied MechanicsCHALMERS UNIVERSITY OF TECHNOLOGYGothenburg, Sweden 2016

MASTER’S THESIS IN AUTOMOTIVE ENGINEERING

Unsteady Thermal Management Simulations for a Passenger Vehicle using1D and 3D Tools

SAKET KUMARDANIEL PASCUAL

Department of Applied MechanicsDivision of VEAS and Combustion

CHALMERS UNIVERSITY OF TECHNOLOGY

Gothenburg, Sweden 2016

Unsteady Thermal Management Simulations for a Passenger Vehicle using 1D and 3D ToolsSAKET KUMARDANIEL PASCUAL

c© SAKET KUMAR, DANIEL PASCUAL, 2016

Master’s thesis 2016:06ISSN 1652-8557Department of Applied MechanicsDivision of VEAS and CombustionChalmers University of TechnologySE-412 96 GothenburgSwedenTelephone: +46 (0)31-772 1000

Cover:Temperature profile across the longitudinal-cross section of the underhood region of a Volvo S80 sedan, parkedin a quiescent environment, obtained from steady state simulation in STAR-CCM+.

Chalmers ReproserviceGothenburg, Sweden 2016

Unsteady Thermal Management Simulations for a Passenger Vehicle using 1D and 3D ToolsMaster’s thesis in Automotive EngineeringSAKET KUMARDANIEL PASCUALDepartment of Applied MechanicsDivision of VEAS and CombustionChalmers University of Technology

Abstract

This work develops a coupled 1D-3D computational tool for analyzing conjugate heat transfer in the enginebay of a passenger vehicle. 1D thermal models of powertrain in conjunction with detailed 3D models of flowstructures can provide a powerful means to correctly capture the physics of the system while keeping thecomputational cost reasonably low. To the authors’ best knowledge, simultaneous coupling of 1D and 3Dcomputational tools has not been applied to modelling thermal interaction in the vehicle powertrain before.Thus, this work constitutes a first step towards performing such coupled simulations and analyzing theiroutcome.

The present study builds on a 1D model of a powertrain cooling system, built in GT-SUITE software,by adding a comprehensive lubrication circuit and mean value engine model to monitor the time dependenttemperatures of oil and engine solids. A 3D model of natural convection & radiation in the engine bay was builtin STAR-CCM+ where the primary boundaries for heat transfer between the surrounding air in the underhoodregion and engine solids were defined. The two models were run simultaneously and data were exchanged forthe surface temperatures of the primary engine components as well as the heat transfer rate between thesecomponents and the engine bay surroundings.

The developed tool was used to study the variation of mass averaged oil temperature over a customizeddriving cycle. This cycle consists of the recently introduced Worldwide harmonized Light duty driving TestCycle (WLTC) with additional shut-down periods during which the car was assumed to be parked in a quiescentenvironment. The delivered numerical procedure proved to be an effective approach for such drive cycles. Thepresent work can thus serve as a solid basis for implementing more advanced thermal management models inthe future.

Keywords: coupled 1D-3D computational tool, conjugate heat transfer, GT-SUITE, mean value engine model,natural convection, radiation, STAR-CCM+, mass-averaged oil temperature, WLTC

i

ii

Preface

This master thesis has been completed in partial fulfilment of the requirements for a Master’s degree inAutomotive Engineering. It is a cross-divisional project between the Vehicle Engineering & AutonomousSystems (VEAS) and Combustion divisions of the Applied Mechanics department at Chalmers University ofTechnology, Gothenburg, Sweden. The primary work has been performed from January to June 2016 by twostudents in the division of VEAS while utilizing resources from both the participating divisions.

Acknowledgements

We would like to express our utmost sincere gratitude to all those who have helped us through the differentstages of this project. First and foremost, we would like to thank our project supervisors, Blago Minovski andJelena Andric. Without their efficient guidance, critical inputs and untiring support, this project could neverhave been completed. We would also like to thank them for providing us with all the required resources andsetting-up a comfortable working environment which helped in the smooth operation of our project. It washighly motivating and a great pleasure for us to work under their supervision.

Furthermore, we would like to thank the engineers of Volvo Cars Corporation of Gothenburg for all thedata provided by them, especially Mirko Bovo from their Engine CAE department, for his valuable insightsand feedback on our work.

A special thanks goes to our examiners at Chalmers, Prof. Lennart Lofdahl and Prof. Sven Andersson fortheir active interest and belief in our efforts as well as the spontaneous humour whenever required.

Last but not the least, we would like to thank all the support staff at Gamma Technologies and CD-adapcofor their advice and technical inputs throughout our project.

Saket Kumar and Daniel Pascual

Gothenburg, September 12, 2016

iii

iv

Contents

Abstract i

Preface iii

Acknowledgements iii

Contents v

Nomenclature viii

Abbreviations viii

1 Introduction & Motivation 11.1 Project Background and Scientific Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Thesis Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Theoretical and Modelling Background 42.1 Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.1.1 Conductive Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.1.2 Convective Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.1.3 Radiation Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.2 Governing Equations of Fluid Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.3 Turbulence Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.4 Grashof Number (Gr) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.5 Cooling Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.5.1 Coolant Pump . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.5.2 Thermostat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.5.3 Charge Air Cooler(CAC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.5.4 Radiator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.5.5 Fan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.5.6 Expansion Tank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.5.7 Lubrication System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.5.8 Oil Cooler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.5.9 Driving Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.6 Computer modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.6.1 GT-SUITE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.6.2 Artificial Neural Networks (ANN) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.7 Macro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.8 Comma Separated Values File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3 Methods 143.1 Modelling Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143.2 3D CFD Model of Underhood Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153.2.1 Generated Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163.2.2 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173.2.3 Solver Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.2.4 Results from Standalone Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.3 1D Model of powertrain cooling system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.3.1 Upgrading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.3.2 Oil Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.3.3 Mean Value Engine Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.4 Final Coupled Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293.4.1 Coupling Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

v

3.4.2 Quasi-Stationary Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.4.3 Coupling Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.4.4 Customized Drive Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.4.5 Ambient Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4 Results and Discussion 344.1 Predicted Oil Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344.2 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344.3 Thermal Soak Phenomenon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

5 Concluding Remarks 37

6 Future Work 38

Bibliography 38

A Appendix IIA.1 Java Macro Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IIA.2 Distribution of Content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX

vi

LIST OF FIGURES

1.1 Subject vehicle: Volvo S80 sedan; source: google.com . . . . . . . . . . . . . . . . . . . . . . . . 22.1 Pump of the cooling system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2 Thermostat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.3 Charge air cooler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.4 Radiator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.5 Vehicle fan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.6 Expansion tank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.7 Schematic of Primary Heat Exchanges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.8 WLTC Class 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.9 GT user interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.10 Artificial Neural Network of two inputs, three neurons and one output . . . . . . . . . . . . . . 123.1 Schematic representation of the all the interactions taking place between different elements of

the entire modelling domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143.2 Division of modelling domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153.3 Computational domain for the 3D-CFD model defined in STAR-CCM+ . . . . . . . . . . . . . 153.4 Longitudinal cross-sectional view of the generated computational mesh grid . . . . . . . . . . . 173.5 Boundaries defined for the exchange of information between the two software . . . . . . . . . . 183.6 Temperature profile across the longitudinal-cross section obtained from steady state simulation

with nominal boundary conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.7 Detailed Lubrication Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.8 Reynolds Vs Loss coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.9 Areas related to the piston lubrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.10 Bottom area of the engine block . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.11 Cylinder head surface measured . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.12 Simplified Oil Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.13 Complete Thermal Engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.14 Comparison between experimental and computed data . . . . . . . . . . . . . . . . . . . . . . . 263.15 Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273.16 Comparison of the brake power computed by the mean value model and the detailed gas exchange

model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.17 Comparison of the exhaust manifold temperature computed by the mean value model and the

detailed gas exchange model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.18 Comparison of the exhaust manifold temperature computed by the mean value model and the

detailed gas exchange model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293.19 CosimFileBased template in GT-SUITE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.20 Schematic of the coupling strategy employed by the Java macro for the exchange of data between

the two software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.21 Customized drive cycle constructed for analysis of the coupled simulation tool . . . . . . . . . . 334.1 Variation of temperature of oil in the sump for simulation number discussed in section . . . . 344.2 Mass-averaged oil sump temperatures from coupled simulations . . . . . . . . . . . . . . . . . . 354.3 Mass-averaged exhaust manifold temperatures from coupled simulations . . . . . . . . . . . . . 364.4 Thermal soak phenomenon observed for the cylinder head . . . . . . . . . . . . . . . . . . . . . 36

vii

Nomenclature

Parameter Symbol Value UnitArea A m2

Convection heat transfer coefficient α0 Wm−2K−1

Engine speed N RPMInternal energy U JLower heating value Qhv kJkg−1

Mass m kgMass flow of air ma kgs−1

Thermal conductivity k Wm−1K−1

Time t sSpecific heat capacity cp Jkg−1K−1

Stefan-Boltzmann constant σ 5.67 ∗ 108 Wm−2

Surface area As m2

Surface temperature Ts◦K

Rate of heat transfer Q WRate of heat transfer per area q Wm−2

Abbreviations

CAC Charge Air cooler

HTC Heat transfer coefficient

WLTC Worldwide harmonized light vehicles test cycle

PCS Powertrain cooling systems

viii

1

Introduction & Motivation

1.1 Project Background and Scientific Relevance

In the current scenario of constantly rising fuel efficiency targets, a lot of research work being done in theautomotive industry is devoted to developing vehicle thermal management solutions [1][2] that can contribute inthis regard. Employing thermal encapsulations for the engine is a technique for reducing the heat it dissipates tothe environment, after it has been switched off. This approach is beneficial for the reduction of fuel consumptionand gaseous emissions. By encapsulating the engine, the oil temperatures are maintained at much higher levelstill the next instance of running of the vehicle. In this way, it is possible to avoid the cold start-up periods andhence minimize the friction losses & exhaust emissions. The efforts of a major automobile manufacturer in thisdirection were recently recognized by the European Commission recently [3]. In order to implement solutionswith similar capabilities, it is essential to understand and model the conjugate heat transfer phenomenonin the engine bay of parked vehicles. Experiments have been carried out to improve the understanding ofbuoyancy driven flows in the engine bay of vehicles that are parked in a quiescent environment after havingbeen subjected to a typical drive cycle [4].

At the same time, there is a rising demand for developing simulation models to minimize or even eliminatephysical testing. These help significantly in reducing the cost and development life-cycle time of the vehicles.The solutions for vehicle thermal management have a significant effect on the overall architecture of a vehicle.Therefore the ability to perform reliable thermal management simulations at concept stage is critical foraccelerated development times. For a vehicle parked in a quiescent environment, buoyancy driven convectionand radiation are the dominating modes of heat transfer in the underhood region. A well-documented numericalmethod described by Minovski et al.[5] offers a reliable procedure for modelling this phenomenon on a commercial3D CFD platform. On the other hand, 1D simulation tools are typically favoured for transient simulations ofpowertrain cooling systems. This work investigates the possibility of coupling these 1D and 3D simulationplatforms to correctly predict the conjugate heat transfer in the engine bay while reducing computational costs.It constitutes the first step towards performing simultaneously coupled simulations of this nature.

1.2 Thesis Scope

This thesis work comprises the following major steps :

1) Further development of an integrated 1D GT-POWER model of the powertraing cooling system of VolvoS80. The model is upgraded by adding the oil circuit elements to the engine thermal model. A mean valueengine mode is implemented to more accurately capture the gas-exchange process.

2) Creation of a 3D CFD model of the vehicle underhood flow using a commerical CFD software STAR CCM+. This model predicts natural convection heat transfer in the engine bay based on the method previouslyproposed by Minovski et al.[5].

3) Coupling between the 1D and 3D simulation tools in order to capture the dependence between the buoyancydriven flow in the underhood region during the thermal soak period and the heat transfer in the engine coolantand lubrication circuits.

4) Performing coupled 1D-3D simulations using a customized driving cycle to investigate the feasibility of thecoupled modeling framework.

1

Figure 1.1: Subject vehicle: Volvo S80 sedan; source: google.com

2

1.3 Limitations

The study has been carried out under the following limitations:

1)The simulation results will be examined for consistency from a theoretical perspective. However, acomparison with experimental date is beyond the scope of the present work.

2) The heat exchange due to forced convection during the periods when the engine is turned on are neglected.However, the primary effect of this flow are taken into account by assuming appropriate values for the heattransfer rate available in the literature.

3) Due to limited time frame, the 3D CFD simulations consider the heat dissipation in the engine bay to be aquasi-stationary process, refer to Sect. 3.4.2 for more details.

4) A uniform temperature distribution has been assumed across the engine surfaces.

3

2

Theoretical and Modelling BackgroundThis chapter introduces the readers to the theoretical background of this project. To begin with, it highlights theprimary concepts on which the models are based. A brief description of the primary components of powertraincooling systems follows and the topics of computer aided modelling relevant to this work are introduced here aswell.

2.1 Heat Transfer

Matter is made up of molecules and atoms that are in constant motion and possess heat or thermal energy.The greater the particle motion and energy levels, the higher will be the temperature of the system. Heattransfer always occurs from a higher-temperature system to a lower-temperature system. The different modesof heat transfer are conduction, convection and radiation. The respective characteristics and concepts behindsthese are discussed below.

2.1.1 Conductive Heat Transfer

Conduction is the phenomenon that consists of heat propagation between two bodies or different parts fromthe same body that are at different temperatures due to the agitation of thermal molecules.

The Fourier’s Law for heat conduction in one-dimensional form is given as:

Qcond = −kAdTdx

(2.1)

where Qcond heat flow through a cross-sectional area A and k is the thermal conductivity. [6].

2.1.2 Convective Heat Transfer

Convection is defined as the transmission of heat resulting from the bulk motion of the molecules. Thisphenomenon only occurs in fluids as their particles can exhibit bulk current flows and significant diffusion.These can occur due to density differences (natural convection) or by external means such as pumps andfans(forced convection).Convective heat transfer is mathematically described by the equation 2.2 [6]:

Qconv = hA(Tinf − Ts) (2.2)

It can be deduced from the above equation 2.2 establishes that the heat flux Qconv between a fluid attemperature Tinf measured far away from the wall that is at temperature Ts is proportional to the area incontact h,heat transfer coefficient, and the temperature difference between both.

2.1.3 Radiation Heat Transfer

This mode of heat transfer exists between any two bodies that are at different temperatures even without anyphysical connection or material medium between them. Radiation is a form of electromagnetic emission that isemitted by any body which is at a temperature higher than absolute zero. The emission of an idealized blackbody can be calculated with equation 2.3 [6]:

Qrad = σAsTs4 (2.3)

Where As and Ts are the area and the temperature of the black body respectively and σ = 5.67 ∗ 108Wm−2

is the Stefan-Boltzmann-constant. For a non black body, this phenomenon is described by the equation 2.4 [6]:

Qrad = εσAsTs4, (2.4)

where ε is the emissivity of the surface that also depends on the temperature of the body and varies withinthe range 0 < ε < 1 [6].

4

2.2 Governing Equations of Fluid Flow

The governing equations for incompressible flow comprise of three main equations. These are the equations ofmass continuity, conservation of momentum and the conservation of energy respectively. They are given as inequations 2.5,2.6 and 2.7 below:

∂ui∂xi

= 0 (2.5)

∂ui∂t

+ uj∂ui∂xj

= −1

ρ

∂P

∂xi+ ν∆ui (2.6)

∂e

∂t+ uj

∂e

∂xj= φ+

1

ρ

∂(K ∂T∂xj

)

∂xj(2.7)

where ui is the velocity component in the i-th direction, p is the pressure, ρ and ν denote the density andthe kinematic viscosity of the fluid respectively. Also ∆ is the Laplacian operator.e is the internal energy per unit mass and φ is the viscous dissipation per unit mass given as:

φ =ν

2(∂ui∂xj

+∂uj∂xi

)(∂ui∂xj

+∂uj∂xi

) (2.8)

The governing equations are formulated for the instantaneous flow field.

Reynolds decomposition technique is used to account for the equation terms for turbulent flows. The in-stantaneous variables are represented as a sum of their time-averaged and fluctuating components. For theinstantaneous velocity and pressure fields, this decompositions reads as:

ui = Ui + u′i (2.9)

p = P + p′ (2.10)

Ui and P are the time averaged components, and u′i and p’ are the fluctuating components.The Reynolds averaged Navier Stokes (RANS) equations are given as:

∂Ui∂xi

= 0 (2.11)

Uj∂Ui∂xj

=∂[−P

ρ ∂ij + 2νSij − u′iu′j ]

∂xj(2.12)

Here ∂ij is the is the symbol for kronecker delta function;

Sij =1

2(∂Ui∂xj

+∂Uj∂xi

) is the mean strain tensor (2.13)

The term −u′iu′j in equation 2.12 is referred to as the Reynolds stress. This non-linear term needs to bemodelled in order to solve the RANS equations.

5

2.3 Turbulence Modelling

The presented work employs the Realizable k-ε turbulence model. In this model, the approximation for theReynolds stress is based on the Bousssinesq assumption, i.e.

− ρuiuj = 2µtSij −2

3K∂ij (2.14)

here µt is given as: µt = ρCµK2

ε

where K is the turbulent kinetic energy and ε is the turbulent dissipation. Contrary to the standard k-εturbulence model, Cµ is considered a variable in this approach, see [7] for more details. Realizable k-ε modelhas better ability to correctly capture the physics of buoyancy driven turbulent flows compared to the otheroptions such as the standard k-ω, SST k-ω etc.

2.4 Grashof Number (Gr)

The Grashof’s number (Gr), named after the German engineer Franz Grashof, is a dimensionless quantity usedfor analyzing the flow characteristics and the velocity distribution in scenarios where buoyancy driven flowexists. The turbulent nature of the buoyancy driven flow in the engine bay of a vehicle can be identified withthe help of this quantity. Physically, it is the ratio of the buoyant forces to the viscous forces and is analogousto the Reynolds number for cases of forced convection. Mathematically,

Gr =Buoyantforces

V iscousforces(2.15)

Also,

GrL =gβ(Ts − T∞)L3

ν2; for vertical flat platesfor vertical flat plates (2.16)

and

GrD =gβ(Ts − T∞)D3

ν2; for bluff bodies (2.17)

where:g is acceleration due to Earth’s gravityβ is the coefficient of thermal expansion (equal to approximately 1/T, for ideal gases)Ts is the surface temperatureT∞ is the bulk temperatureL is the vertical lengthD is the diameterν is the kinematic viscosity.

while The L and D subscripts indicate the length scale basis for the Grashof Number. In the cases ofbuoyancy driven flow from flat vertical plates, the transition to turbulent flow is characterized by the occurrenceof Grashof number in the range specified as 108 < GrL < 109. For higher values of this number the boundarylayer is turbulent while at lower Grashof numbers, it is laminar.

The product of the Grashof number and the Prandtl number gives the Rayleigh number, which is anotherdimensionless number that characterizes convection problems in heat transfer. The theory governing buoyancydriven flows is available in more detail in the dedicated literature works in this field.[7].

The following section introduces the readers to the primary components of a vehicle powertrain coolingsystem that have been modelled in this project.

6

2.5 Cooling Systems

The main aim of all cooling system is to maintain the operational temperatures of the powertrain within desiredlimits and prevent it from reaching any critical scenarios where the components may get damaged. For internalcombustion engines, initial warming up of the engine during cold starts, cooling down of pistons, engine blockand cylinder head are the main functions of the cooling system. In order to carry out the cool down there aretwo ways to do it, either using air directly as cold medium or using a coolant liquid as intermediate medium tocool down the engine and then cooling down the coolant liquid using a radiator.

2.5.1 Coolant Pump

The coolant pump is the device in charge of supplying energy to coolant fluid in order to circulate it throughthe coolant circuit.Pump usually is connected to the crankshaft using a belt with a speed ratio, from thisconnection extracts the necessary energy to rotate. A sketch of centrifugal pump [8] can be found on Figure 2.1.

Figure 2.1: Pump of the cooling system

2.5.2 Thermostat

A thermostat [9] is a valve which controls coolant flow going from engine to the radiator. There are two differenttypes of the thermostats, first type is based on wax tensibility. Second type of thermostats is based on the fluidmixture tensibility with a mixture of water and alcohol. Nowadays, the most common thermostat is that of thewax type. When the thermostat is opening, the coolant starts to flow to the radiator circuit. The thermostatopens depending on the coolant temperature and when the temperature of the coolant reach 70–80◦C starts toopen. The thermostat is fully opened after reaching the temperature 85–95◦C. These values can vary dependingon the car and the strategy followed. [10]

Figure 2.2: Thermostat

7

2.5.3 Charge Air Cooler(CAC)

Forced induction is an effective method for improving the power output of internal combustion engines whilemaintaining their displacement or size. Charging of the engines is provided by the turbocharger or supercharger.However, increasing the temperature of the compressed air is not useful to rise the engine power. In order toenhance the efficiency of the combustion this air coming from the compressor is cooled down in the Charge aircooler [11] before going to the combustion chamber to increase its density and obtain a powerful combustion[10].

Figure 2.3: Charge air cooler

2.5.4 Radiator

A radiator is an liquid to air, cross-flow type heat exchanger that usually is placed after charge air cooler. It isthe principal effective interface for heat exchange between coolant and surroundings environment. The mostcommonly used design for modern vehicle radiators is the finned aluminium type that uses flat tubes held byheader plates. An example of a passenger car radiator [12] is shown on Figure 2.4

Figure 2.4: Radiator

The thickness and face areas sizes of the radiator cores can vary. Pressure drop across the coolant side ofthe radiator is mainly affected by the number and dimensions of the tubes where the liquid flows. At the airside is fin geometry what has high influence on the drop pressure and cooling efficiency of the radiator. Heat istransferred from the hot coolant to the wall by convection, by conduction through the wall and to the coldfluid by convection again.

Proper selection of the radiator is a very important task in automotive engineering that requires large

8

amounts of computer simulations to reach the requirements and design targets. Nevertheless it is necessary tohave previous performance knowledge of the radiator such as pressure loss and cooling rate that usually areprovided by the supplier [13].

2.5.5 Fan

In order to increase air flow through the radiator at low speeds a fan is installed. Depending on the size ofthe engine combination of two and more fans can be used. Fans [14] can be driven by the engine, howevernowadays fan speed is controlled by electric motor,making it independent to engine. speed.

Figure 2.5: Vehicle fan

2.5.6 Expansion Tank

Expansion tank is a solid or deformable tank mounted at the top of the cooling system. The function of thistank is to permit coolant expansion at high temperatures since if the coolant is not allowed to expand, thepressure reached in the coolant circuit would be too high.(see Fig. 2.6) [15].

Figure 2.6: Expansion tank

9

2.5.7 Lubrication System

The lubrication system is responsible for friction reduction between moving parts inside the engine as wellas protect them from wear and tear. If the lubrication layer breaks down, components in motion can reachtemperatures high enough to lead to localized welding and hence failure of the engine. The lubricant is alwaysin contact with the most critical components of the engine since its objective is to reduce friction, so at thesame time that reduces friction levels,it extracts heat generated because of the inevitable friction and fromthe combustion process. A clear example of this double purpose is the piston lubrication throughout cylindermovement. Before being sent back to the sump, the hot lubricant is cooled down through the oil cooler . Aschematic of primary heat exchanges in an engine is depicted in the Figure 2.7

Figure 2.7: Schematic of Primary Heat Exchanges

2.5.8 Oil Cooler

Lubrication of the different moving components of an engine is essential at every moment. It is used forreducing friction losses at the engine. But the fluid used to this purpose, oil, gets warm up after lubricatingand it important to maintain the oil at a temperature lower than certain value and it can keep its properties.The purpose of a oil cooler is to maintain the temperature of the oil under certain value since if it reaches hightemperature, the properties of the oil would change and the lubrication of the engine would not be proper.

Nevertheless the lubricant can be also warmed up in the oil cooler, for instance, at a cold start, the coolantwarms up faster than the oil leading to a temporary phase where the oil is warmed up by the hotter coolant inthe oil cooler. This helps in decreasing oil warm-up period and therefore reducing the friction losses due to thegreater viscosity of the oil at low temperatures.[10]

10

2.5.9 Driving Cycle

A driving cycle may be defined as a series of vehicle speed values, given as a function of time for a specifiedduration. Driving cycles are used for quantifying the performance of vehicles on the basis various parameterssuch as fuel consumption and polluting emissions in a normalized way, so that different vehicles can be compared.They are developed by automotive experts across the globe to represent naturalistic driving patterns. Examplesof driving cycles include the New European Driving Cycle (NEDC) and the more recently introduced Worldwideharmonized Light vehicles Test Cycle (WLTC), which is illustrated in figure 2.8 below:

Figure 2.8: WLTC Class 3

2.6 Computer modeling

This section introduces the readers to the modelling platforms, techniques and other computing details relevantto this project.

2.6.1 GT-SUITE

GT-SUITE is a multiphysics platform for simulating different vehicle technical areas. This platform allowsintegration of different tools under a common interface that can be simulated simultaneously. For detailedinformation about GT see reference [16].

The modeling in GT-Suite is template based and the user interface used is called GT-ISE. In the Figure 2.9the templates can be found in the left sided library and after picking the template and define it becomes aobject. This object can be dragged into the model map. One of the advantages of GT-Suite is that a singleobject can be used to create other objects with similar attributes. This can be done by using different names,but all these objects will be under a ”parent” object. Thus, anything changed in the ”parent” object will affectthe attributes of these objects. Once these objects are created and dragged to the model map they can beconnected with other different objects to create a flow circuit for example the coolant flow circuit shown in theFigure 2.9. Some of the options in the object can be parametrized. This can be done by assigning a name forthis parameter in brackets. [10]

2.6.2 Artificial Neural Networks (ANN)

Artificial neural networks are automated processes and a learning paradigm and inspired from the humanbiological nervous system. They consist of an interconnected system of neurons that share information amongthem to produce an output. Each neuron receives a certain number of inputs and sends an output dependingon two functions, namely: input function 2.18 and activation function 2.19.

11

Figure 2.9: GT user interface

s =

N∑j=1

wjxj + b, (2.18)

where N is the number of inputs of each neuron, x the input value, w is the weight of each input and b is thebias, which is a constant value used in case all inputs are zero.

Figure 2.10: Artificial Neural Network of two inputs, three neurons and one output

σ(s) =1

1 + egs, (2.19)

where g is a constant.The ANN reads the input values and output values to train itself using an iterative method to compute the

indicated weights of each input until a certain minimum error value is achieved between the predicted and thetarget output values. The weights are optimized by backward propagation of the error during training. Anexample of a basic ANN is shown in the Figure 2.10 [17]

12

2.7 Macro

A macro or a ”macroinstruction” is essentially a fragment of programming code that contains a sequence ofinstructions to be executed in order to complete certain routines. The macros for STAR-CCM+ are written instandard Java programming language which enables the user to access to all the programming constructs of thatlanguage, such as loops and conditional constructs. They are compiled and executed within the STAR-CCM+workspace which enables the user to establish two-way communication with any external software for exchangingthe values of different parameters. It is highly recommended to use an integrated development environment oran ”IDE” for compiling, writing and debugging such macros as they help the users to identify the potentialerrors in the code. The NetBeans IDE 8.1 has been utilized for the purposes of this project since it is fullycompatible with all the instruction libraries of STAR-CCM+ and is recommended by CD-adapco.

2.8 Comma Separated Values File

A Comma Separated Values (CSV) file is a type of delimited text file that stores tabular data (numbers andtext) in plain text. Each of the successive fields in a CSV file are separated from the former with a comma (,).CSV formats are best used to represent sets or sequences of records in which each record has an identical list offields. This corresponds to a single relation in a relational database, or to data (though not calculations) in atypical spreadsheet. The format is very popular in the field of computing and is widely used to pass data withdifferent internal word sizes, for data formatting needs, and so forth. For this reason, CSV files are common onall computer platforms.

13

3

MethodsThis chapter introduces the readers to the different methods and procedures utilized in this project. It beginswith giving an overall view of the various interactions taking place within the entire domain that this projectaims to capture. This is followed by the description of the methods adopted for modelling of the 3D and 1Dmodels respectively.

3.1 Modelling Domains

As a first step, the different interactions involving the transfer of mass and thermal energy taking place in thephysical environment of our interest, which is the entire vehicle underhood region, are identified and studied.These interactions are depicted in the form of a schematic representation in figure 3.1 below. This studyhighlighted that the phenomena and systems whose modelling is of primary interest for this project are:

• Buoyancy driven flow in the underhood region of the subject vehicle when it is parked in a quiescentenvironment after being subjected to a driving routine.

• Its powertrain cooling systems and the various thermal interactions among them.

• Conjugate heat transfer between the external surfaces of the engine solids and their surrounding space inthe underhood region.

Figure 3.1: Schematic representation of the all the interactions taking place between different elements of theentire modelling domain

Based on their respective nature and complexity, the modelling of these phenomena and system is separatedinto two major domains where 1D and 3D CFD modelling are required respectively. The buoyancy driven flowin the underhood region is unsteady and turbulent in nature. The properties of the resulting plumes and flowstructures are inherently 3D in nature. High spatial and temporal resolution are required for their analysis[?].Hence a 3D CFD model, built in a commercially popular CFD package called STAR-CCM+ (Version 10.06.009)is employed for capturing the physics of this phenomenon.

Powertrain cooling systems (PCS) consist of a multitude of components whose functioning and mutualinteractions are governed by a number of parameters. Transient analysis of these systems essentially requires1D modelling in order to keep the computational demands within realistic limits. A previously built 1D modelfor the subject vehicle PCS is upgraded and utilized for their analysis. The commercial software GT-SUITE(Version 7.5) is used for the development of this model. The main functionality of this model consists of runningtransient simulations with the VED5 engine of the subject vehicle and its powertrain cooling systems (PCS)thereby capturing all the relevant thermal interactions taking place among them.

14

Finally to capture the conjugate heat transfer (CHT) between the systems represented in these two models,mutual coupling is required to link the data processed by them individually. This aspect of the project will befurther elaborated ahead in section 3.4. Hence the division of the physical systems modelled in this project canbe summarized in the form of a block diagram in figure 3.2.

Figure 3.2: Division of modelling domains

3.2 3D CFD Model of Underhood Region

A detailed surface mesh of the complete vehicle provided by Volvo cars is available as an input to this project.Primary preprocessing of this mesh has been done in ANSA to define a computational volume illustrated infigure 3.3. This volume comprises of a cuboidal enclosure of dimensions measuring approximately 2 m(length)X 2 m(breadth) X 1.6 m (height).

Figure 3.3: Computational domain for the 3D-CFD model defined in STAR-CCM+

15

It captures the frontal half of the vehicle extending from the grille to the A-pillars, making sure thatthe entire underhood region and its immediate surroundings are included. This initial representation of thecomputational volume is then exported to STAR-CCM+. Since the heat transfer within the engine solids andamong the powertrain cooling system components is modelled in the 1D model in GT-SUITE, the fluid domainin the underhood region was of primary interest for this model to capture the convective and radiative modesof heat transfer.

3.2.1 Generated Mesh

The region based meshing approach is employed for this model since it is a tried and tested method for suchapplications [?]. To mesh the fluid domain of the computational volume, the following meshing models inSTAR-CCM+ are employed:

• Surface Wrapper: Given the presence of multiple intersecting surfaces and gaps in the provided rawmesh, the generation of a well-resolved, water tight surface mesh is required as a first step. The surfacewrapper model is used for this purpose as it generated the desired mesh much faster and easily comparedto the wrapper available in ANSA, and has been recommended in several published works [?]. It alsoprovides the option of including refinements that are based on curvature, proximity, and individualsurfaces out of which the curvature refinement option is utilized in this case. However the resultingsurface quality from this model is not optimal. Therefore, it is used with the surface remesher model, aselaborated ahead, to provide a high quality starting surface for the volume mesher.

• Surface Remesher: To improve the overall quality of the surface mesh generated using the wrappermodel and to existing surface and optimize it for the volume mesh models, the surface remesher is beused to re-triangulate the surface for improving its overall quality and optimizing it for the generation ofthe volume mesh. The target edge length is specified as 5 mm along with both curvature and proximityrefinement for capturing the complex geometries of the underhood region. Localized relaxation in themesh size is employed for the fluid region surrounding the vehicle components to reduce the unnecessaryexpenditure of computational resources. The surface growth rate is specified as 1.3. A list of the valuesspecified for the surface mesh is presented in table 3.2 below. The option of aligned meshing is selected inthe remesher model to optimize the resulting surface triangulation further.

Parameter ValueTarget cell size for vehicle component boundary surfaces 5 mm

Minimum cell size 3 mm

Localized target cell size for fluid region 300 mm

Localized minimum cell size for fluid region 250 mmSurface growth rate 1.3

Table 3.1: Specification of surface mesh generated for the fluid domain

16

• Advancing Layer Mesher: The Advancing Layer model, which has been introduced recently bySTAR-CCM+, is utilized for generating the volume mesh. It converts the existent triangular surface meshinto a polygonal surface. This polygonal surface is then advanced into the domain volume by extruding itinto the region volume to form prismatic cell layers. As a result the dependency between the subsurfaceand region topology is reduced which allows it to generate a thicker, more uniform layer than the prismlayer mesher can achieve. 3 layers of prismatic cells are generated around the external surface mesh whichcapture the boundary layer, turbulence effects, and heat transfer near wall boundaries. The prism layerthickness is set to 1.3 mm. The remaining surrounding void of fluid region is filled with polyhedral cells.The advancing layer mesher also collapses and splits faces and edges, which helps to improve mesh quality.The final generated volume mesh comprised of approximately 50 million cells. The longitudinal crosssectional view of this mesh is depicted below in figure 3.4.

Figure 3.4: Longitudinal cross-sectional view of the generated computational mesh grid

3.2.2 Boundary Conditions

Since this model represents the vehicle standing in a quiescent environment, atmospheric pressure is assignedto the fluid domain. The ambient temperature is set to 20 degrees Celsius. The main interfaces through whichthe conjugate heat transfer takes place between the engine and the surrounding air are defined as illustrated infigure 3.5.

17

These include the external surfaces of the following engine solids:

• Engine cylinder head

• Engine block

• Exhaust manifold

• Oil Sump

Figure 3.5: Boundaries defined for the exchange of information between the two software

In order to verify the functionality of the model before using it for running coupled simulations, certainverification simulations are performed with the 3D model running in isolation. Nominal values for thetemperature are assigned to the boundary surfaces specified above for performing these simulations. Thesevalues are given as:

• Ambient : 20◦

• Engine block : 100◦

• Cylinder Head : 110◦

• Oil Sump : 125◦

• Manifold : 600◦

At the same time, all other surfaces present in the model are modelled as adiabatic components so thatwhile the influence of their geometrical detail on the flow structures is captured, their participation in thethermal interactions taking place in this model is neglected for simplicity. In the final coupled model, thetemperature values for these boundaries will be assigned through the data obtained from the 1D model inGT-SUITE in real-time.

18

3.2.3 Solver Settings

The major setting employed in this model for the solver are enlisted below:

• Steady state solver: Owing to the complexity and the number of cells in the computational grid inthis model, the steady state solver is employed. Transient simulations with models of such detail arecomputationally very expensive.

• Fluid Selection: All the physical and chemical properties of air are assigned to the fluid surroundingthe powertrain components in this region. Ideal gas laws are used for modelling the governing physics forthis fluid.

• Gravity and Radiation: To effectively capture the flow dynamics and the modes of heat transfer,gravity and gray-thermal radiation models are included. The size of the radiation patches are set to aface proportion of 30 percent, implying that that the mesh cell size is approximately 30 percent of thepatch size.

• Realizable k-ε Turbulence Model: STAR-CCM+ offers three different approaches for the definitionof turbulent eddy-viscosity, namely: Spalart-Allmaras, k-ε and k-ω models. A linear relationship betweenthe strain-rate and the Reynolds-stress tensor, better known as Boussinesq assumption, is assumed inall of them with the proportionality referred to as ”turbulent eddy-viscosity”. Each of these model aresuitable for specific applications. The standard k-ε model offers a good compromise between robustnessand accuracy for a larger range of flow configurations. Compared to other options such as the standardk-ω and the SST k-ω model, it does not have the tendency to overpredict the turbulence levels in regionswith large normal strain, like stagnation regions and regions with strong acceleration, which is existent inthe boundary layer around the hot engine solids. The realizable k-ε model is substantially better thaneven the standard k-ε model and can generally be relied upon to give results that are at least as accurate.

• Wall Treatment: The k-ε model is implemented in STAR-CCM+ with a Two-Layer All y+ walltreatment, which activates it to be used with fine meshes that resolve the viscous sub-layer. This walltreatment is a hybrid approach that attempts to emulate the high y+ wall treatment for coarse meshesand the low y+ wall treatment for fine meshes and is also formulated in order to produce reasonablesolutions for meshes of intermediate resolution, i.e. when the wall-cell centroid falls within the bufferregion of the boundary layer, where 5 < y+ < 30 .

19

3.2.4 Results from Standalone Simulations

The steady-state simulations with the 3D model achieved convergence after approximately 4000 iterationswhile utilizing a total of around 3000 core-hours each. The resultant temperature field around powertraincomponents showed high temperature distribution in the regions above the exhaust manifold and cylinder headas the plumes rising from these hot surfaces spread around in the engine bay. This distribution is illustrated inthe figure 3.6.

Figure 3.6: Temperature profile across the longitudinal-cross section obtained from steady state simulationwith nominal boundary conditions.

3.3 1D Model of powertrain cooling system

The 1D model of powertrain cooling systems available as an input to this project has been built by a 6-memberproject group during the autumn semester of 2015 at Chalmers University of Technology. It consists of:

• Simplified thermal engine model

• Complete powertrain cooling system

• Mechanic vehicle model

3.3.1 Upgrading

To meet the targets for this thesis work, some necessary changes and upgrades have been made to the givenmodel.

Convergence

The first upgrade carried out was the increase in number of time steps converged in the GT-Solver. Whenthe first simulation was run for the given model using the WLTC (Worldwide harmonized Light vehicles TestCycle) some convergence problems occurred. This is explained because the drive cycle used in this work, WLTCdrive cycle has a higher level of fluctuations in the speed compared to the sample drive cycle provided for thepurposes of that project. One of the most critical parts of the cooling system is the pump since at high engine

20

speed a great amount of energy is transmitted to the fluid. Because of this the relaxation factor of the pumpwas reduced ten times. This change helped to make almost 100 % of the time steps converge but with longercomputational times as the relaxation factor is a parameter used to dampen the solution in iterative solversand cut out oscillations.

Thermal Nodes

The thermal masses of the given thermal engine model are modelled using ThermalMassOld templates inGT-suite to represent the different parts of the engine such as the main block, cylinders and the cylinder head.This template is obsolete and has limited connectivity. All ThermalMassOld objects were replaced by theirsuccessor ThermalNode objects that are basically the same but allow a greater connectivity option for the user.

3.3.2 Oil Circuit

Lubrication Detailed Model

A detailed and complex lubrication model was given by Volvo. This model was designed to compute accuratelythe pressures over all bearings in the engine and its lubrication.

Figure 3.7: Detailed Lubrication Model

Simplification

Because of the complexity of the model, the required simulation time was excessive. As a first step to overcomethis hurdle, the number of flow splits have been reduced all redundant objects were either eliminated or replaced.However it is important keep the mass flow of each branch of the oil circuit in order to know the amount of oilthat will be in contact with the hot parts in the engine.

To achieve this aim, the PressureLossConn template is utilized to impose pressure drops and remove eachof the bearing in the detailed model. This template provides the option to permit the creation of a functionrelating the Reynolds number for the flow through the circuit and the desired pressure drop across it.

The whole lubricated model has been simplified into seven branches that can be classified by:

• Main Bearing branch number 1

• Main Bearing branch number 2

• Main Bearing branch number 3

• Cylinders branch

21

• Turbocharger branch

• Exhaust Camshaft branch

• Intake Camshaft branch

As mentioned above, it is necessary to maintain the flow rates, and thus the behaviour of the circuit iscaptured at diverse working points by running several cases. Pump speed and the temperature of the wall ofpipes are primary parameters that are modified in each of these cases.

All pressures, densities and mass flows at the beginning of each branch are stored to compute the pressureloss coefficient according to equation given below:

∆P =1

2Kρ

(m

ρA

)2

(3.1)

∆P = difference in total pressure across the connection (total upstream - total downstream)K = loss coefficient defined in this templatem = mass flow rate through the ’PressureLossConn’A = reference areaρ = density of the fluidSince all branches of the circuit end at the oil sump ,in our case the value of ∆P is the difference between

the pressure at the beginning of the branch and the atmospheric pressure and the reference area is the sectionarea of the pipe. At the same time the loss coefficient was computed, the Reynolds number of to the branchwas determined in order to create a relation between both values. This procedure was carried out for each ofthe branches. For instance the scatter obtained in the main bearing branch number 1 is 3.8

Figure 3.8: Reynolds Vs Loss coefficient

In order to make the model faster, some pipe objects were simplified to reduce the number of pipe objectswhile maintaining the total length of the circuit. It is important to mention why loss coefficient method waschosen among other different ways that were available to impose a pressure drop. The main advantage of thismethod is that the PressureLossConn template can take into account the temperature of the fluid crossingthrough it, being more specific the density and viscosity of the fluid. It is known that viscosity and density offluids change with the temperature and oil is no exception to this. Since the oil flows at distinct temperaturesthrough the lubrication circuit, the pressure drop due to the pipes, bearings and nozzles will be also distinct.

The way that PressureLossConn template reacts to these changes is by taking into account the Reynoldsnumber, since by definition Reynolds number measure the dynamic viscosity µ, density ρ and velocity of thefluid.

Re =ρvL

µ(3.2)

22

As a result, once all PressureLossConn were computed and the pipes were simplified, a fast and temperaturedependant model was obtained. This model is able to keep the same mass flow rate as the detailed model for atransient simulation.

Identifying the main heat transfer interfaces

In order to capture the oil warm-up inside the engine three different transfer heat areas where the oil is incontact with the hot parts of the engine were considered. The most relevant of them are these where the oil isin contact with the cylinder liners to provide lubrication . When the piston moves throughout the cylinderliner, it generates a high friction levels that need to be reduced. Pressurized oil is used to overcome this bycreating a thin layer between the piston and the liner. Due to the high temperature of the combustion chamberand friction produced the oil warms up. Two areas related to the piston lubrication are seen in Figure 3.9. Theblue area represents the bottom part of the piston and the orange area represents the surface of the liner. Bothareas were measured and introduced into model in order to compute the heat transferred to the oil.

Figure 3.9: Areas related to the piston lubrication

Another relevant heat transfer area is the bottom part of the engine block. This surface was selected sincethe oil used to lubricate the main bearings is splashed due to the spinning motion of the crankshaft. Since theheat transfer area of this phenomenon is very difficult to select accurately, an estimated area was measured inits place. This area selected is shown highlighted in pink in the Figure3.10.

The third heat transfer area considered was the base of the cylinder head. This area was selected since theoil to lubricate the camshaft has to cross through this hot surface before returning to the oil sump and hencethere is a heat transfer between them. The are measured in ANSA is shown in pink in the Figure3.11. Thecylinder head is shown cut in order to see the surface clearer.

The final simplified oil circuit is shown in the Figure3.12. Two different Flowsplit objects were created tocompute the heat transferred by the thermal engine.One represents the area of the Figure3.11 and it is theupper grey box in the Figure3.12 that is connected to the camshaft branches of the oil circuit, and the otheris the grey box situated in the middle of the Figure3.12 which represents the heat transferred through themeasured surfaces 3.9and3.10.

Calibration

The different experimental values of the temperatures achieved in the oil sump are available in the data providedby the Forschungsinstitut fur Kraftfahrwesen und Fahrzeugmotoren institute through the experiments carried

23

Figure 3.10: Bottom area of the engine block

Figure 3.11: Cylinder head surface measured

out by them on Volvo’s behalf. These values have been recorded on a test-rig using the same engine used forthis project. In total 35 cases were recorded in the test-rig varying the velocity of the engine from 750 rpmto 400 rpm and the position of the accelerator from 20% to 100%. The main aim of the calibration of thepredictive oil circuit was to match the same temperature in the oil sump of the model with the experimentallymeasured temperature. For the calibration of the model 35 steady cases with the same boundaries that in theexperimentally cases were run. In order to achieve the same temperature in the predictive oil circuit that inthe experimental data, the value of the heat transfer coefficient 2.2 was modified at each steady case until thedesired temperature was obtained.

The heat transfer coefficient is not a constant value, hence it was necessary to develop a relation betweenthe value of the heat transfer coefficient and the mass flow rate of the oil that was flowing through the oilcircuit since it is a parameter in which the heat transfer coefficient highly relies on. Comparing all the heattransfer coefficient computed for each case in an excel file, a function, namely a power Equation 3.3 was createdto relate the heat transfer coefficient and the mass flow rate through the oil circuit as it was the function thatbetter followed the trend of the values computed at the steady cases.

htc = 1.05712 × 10−5m1.90745, (3.3)

where htc is the heat transfer coefficient and m is the mass flow rate of the oil.The complete oil circuit connected to the thermal engine masses is shown in the Figure 3.13

24

Figure 3.12: Simplified Oil Circuit

Figure 3.13: Complete Thermal Engine

Validation

The validation for transient simulations of the oil circuit has been done by comparing the temperature computedin the oil sump using the predictive oil circuit with experimental data provided by Volvo from a drive cyclewhere the temperature of the oil sump were recorded. As it is shown in Figure 3.14 the model overpredictstemperature in some periods. However, the conclusion obtained from this validation is that it was possible toensure that the oil circuit predicts accurately enough the temperature achieved in the oil sump for transientsimulations.

3.3.3 Mean Value Engine Modeling

The mean value engine is used to model an engine using a map-based cylinder. The mean-value cylinder modeluses maps to characterize the distribution of fuel energy and cylinder air flow. This provides a model thatexecutes more rapidly than a detailed gas exchange model, as the combustion process is not predicted. Meanengine value models are typically used in simulations where computation speed is important and detailed

25

Figure 3.14: Comparison between experimental and computed data

characterization of engine processes is considered relatively unimportant.

There are three parameters entered into a mean value model that are used to characterize the engine cylinderperformance. The first is the volumetric efficiency of the cylinder, that is used to compute the mass flow of airthrough the cylinder that will be imposed throughout the simulation. The second is the indicated efficiency,that is used to quantify how much of the fuel energy contained in the cylinder is turned into work done onthe piston. The final parameter is the exhaust energy fraction, that is used to compute how much of the fuelenergy is employed to warm up the exhaust gases. Then the remainder of the fuel energy that is not used forindicated work or exhaust energy is assumed to be lost through heat transfer [18].

Using a mean value engine instead of a EngineState template which was used to model the engine in theprevious project, a grater engine performance accuracy is obtained at transient simulations. Also the meanvalue engine offers the possibility of calculating the exhaust manifold temperature, which is one of the hottestparts in the underhood and hence must be considered in natural convection cooling.

Design of Experiment Parameters

As mentioned above, the mean value engine is a map-based cylinder so it was necessary to create a cylinderperformance map and a DOE (Design Of Experiment) was prepared. In order to characterize cylinderperformance under numerous situations, a Latin hypercube 1 DOE with 4500 different cases was prepared andthe simulations were run using the detailed gas exchange model provided by Volvo using the following eightparameters with the mentioned ranges:

• Intake manifold pressure [1,2.6](bar)

• Intake manifold temperature [273,450](K)

• Start of injection timing [-26,1](deg)

• EGR valve aperture diameter [0,26](mm)

• Engine speed [750,4500](rpm)

• Exhaust manifold pressure [1,4.6](bar)

• Ambient temperature [273,311](K)

• Fuel injected per cylinder per cycle [0,70](mg)

1Latin hypercube:”It is a method for ensuring that each probability distribution in the model is evenly sampled which at firstglance seems very appealing”[19]

26

Artificial Neural Network

After carrying out all experiments and storing the results, it was necessary to create a look-up process thatpredicts basing on the above inputs each one of the four mean value engine required parameters. To accomplishthis purpose, GT-SUITE provides the option of training a Neural Network, introducing stored data from DOEsimulation as inputs.

Comparing results of three different methods, Self-Organizing Local Linear, Radial Basis Neural Networkand 3-Layer Feed Forward Net and using lower RMS(Root Mean Square) error criteria a 21-neuron 3-LayerFeed Forward Net was trained for volumetric efficiency parameter, a 26-neuron 3-Layer Feed Forward Net forindicated efficiency parameter, a 26-neuron 3-Layer Feed Forward Net for exhaust energy fraction parameterand a 21-neuron 3-Layer Feed Forward Net was trained for the friction mean effective pressure. A diagram ofthe four neural networks is shown in the Figure 3.15

Figure 3.15: Neural Network

Simplifying the Flow System

Because of the mapped nature of the mean value engine model, the flow is steadier than a detailed enginemodel. Consequently the air handling system may be simplified to reduce the simulation run-time. The intakeand exhaust systems may each be lumped into larger equivalent volumes. [18]

Taking these factors into account, all of the pipes before the compressor are simplified into one Flowsplitconserving the total volume and surface. The same procedure is followed for the intake manifold, exhaustmanifold and the pipes before the turbine.

Calibration

The simplification of the air handling system has a couple of effects to be considered. First the pressuredrop reached through the air circuit in the simplified model is lower because of the lack of pipes. This makenecessary to impose a pressure loss to the system through a OrificeConn template with a calibrated diameter.The calibration of the diameter is carried out comparing the maximum pressure drop of the system, which isachieved at the maximum engine load case and matching both values. Second, the heat transferred between thegas and the simplified volumes are less than in the detailed model due to the reduced heat transfer area andthe velocity reached by the flow in the simplified model. This heat transfer difference can be neglected in theintake manifold since the temperature of the air flowing through is not high enough to affect the value of theheat transferred to the wall of the pipes . However the heat transfer in the exhaust manifold was re-adjustedthrough a heat transfer multiplier until the same exhaust temperate was reached.

27

Validation

The validation of the mean value engine was done by comparing values obtained in steady simulations in thedetailed gas exchange model and the values predicted by the mean value engine. The first valued comparedwas the brake power given by the engine. The Figure 3.16 shows the values of the break power obtained usingthe detailed model and the predictive model. Comparing both results it concludes that the predicted brakepower matches satisfactorily.

The second value compared was the temperature of the exhaust manifold. The Figure 3.17 shows as theprevious figure the results obtained using both models. In this case it distinguishes a overpredicted temperatureby the predictive model been the maximum error between both values 9 %.

Figure 3.16: Comparison of the brake power computed by the mean value model and the detailed gas exchangemodel

Figure 3.17: Comparison of the exhaust manifold temperature computed by the mean value model and thedetailed gas exchange model

28

3.4 Final Coupled Model

Along with the individual development and verification of the 1D and 3D models, potential methods for couplingthem are examined and evaluated, leading to the development of the final coupled tool. This section describesthe final stages of development of the coupled simulation tool that is desired as the end-product of this project.

3.4.1 Coupling Mechanism

The two models developed on the 3D and 1D platforms capture the physics and phenomenon occurring in twodifferent domains between which there is no mass transfer.The 3D model of the underhood region captures thevarious physical phenomena that occur for the air surrounding the powertrain components in the engine bay.On the other hand, the 1D model represents the interactions among the powertrain cooling systems themselves.In order to model the conjugate heat transfer in the engine bay, it is essential to capture the interation betweenthe hot surfaces of engine solids and the boundary layer of air surrounding them. This is achieved through theexchange of the values of temperatures and heat transfer rates between them in the following manner:

• First, the temperature values for the engine solids are calculated by the 1D model and imposed on thecorresponding boundaries defined in the 3D model.

• The 3D model then computes the rate at which heat is lost from the surface of these engine solids throughconvection and radiation and feeds their values back to the 1D model.

• The 1D model processes utilizes the data fed from the 3D model and updates the temperatures of theengine solids subsequently. This cycle is subsequently repeated for the entire duration of the simulation.An illustration of the same is shown in figure 3.18.

Figure 3.18: Comparison of the exhaust manifold temperature computed by the mean value model and thedetailed gas exchange model

29

For running co-simulations of the kind required for this project,i.e, where scalar data,i.e the temperatureand heat transfer rates, have to be communicated between two different software, GT-SUITE offers two primarymethods. These are:

• Using an object from the GT-SUITE library, known as CoSimFileBased 3.19. This template enablesco-simulation of GT-SUITE with any third-party software through communication input/output datavia intermediate text files in comma separated values (.csv) format. Two separate .csv files are used totransfer scalar values that are exported from GT-SUITE and imported the external tool.

• Using blocks in simulink to act as the bridge between the two models for exchange of data.

The first option is adopted in this project over the second one. This is done to minimize the complexity ofthe final simulation tool by eliminating the need to have three different software running at the same timeinstead of two. This also reduces the risk of compatibility issues in the future. The first option offered a simplerand more straightforward method while reducing the number of interfaces that had to be set-up. A macro2.7 developed in Java computing language by one of our supervisors, Blago Minovski, is utilized for enablingthe communication from the 3D model’s end. This macro can be used to read and print values of variousparameters in the STAR-CCM+ workspace to any .csv file 2.8 available.

Also, since STAR-CCM+ has not yet adopted the Functional mockup interface standard, hence it is notpossible to establish communication between the models in GT-SUITE and STAR-CCM+ through this methodand hence it is not considered in this regard.

Figure 3.19: CosimFileBased template in GT-SUITE

The data transfer between the different platforms through the described method can be summarized asfollows:

Data exported from GT-SUITE:

• Surface temperatures of engine head, engine block, oil sump and the exhaust manifold

• Engine state, i.e., whether the engine is running or is switched off.

• Elapsed physical time for the simulation.

30

Data exported from STAR-CCM+:

• Heat transfer rate from the surface of the engine head, engine block, oil sump and the exhaust manifoldto their surrounding environment

Data imported by GT-SUITE:

• Heat transfer rate from the surface of the engine head, engine block, oil sump and the exhaust manifoldto their surrounding environment

Data imported by STAR-CCM+:

• Surface temperatures of engine head, engine block, oil sump and the exhaust manifold

The data regarding the engine state is utilized by the Java macro to implement a computational strategyfor the coupled model which will be further elaborated in the section strategysect ahead.

3.4.2 Quasi-Stationary Approximation

The process of heat dissipation from the surface of engine solids in the underhood region is essentially a transientphenomena. The rate at which heat is lost will be a function of time and reduce constantly as the temperatureof engine solids goes down. However, owing to practical reasons relating to available computational resources,the transient solver cannot be used with the 3D CFD model in STAR-CCM+. Hence it is imperative to use analternate method for capturing this process with the coupled tool. The CosimFileBased template in GT-SUITEthat is used for coupling the two models offers the option of controlling the interval of physical time after whichdata is exchanged between them. By employing the steady state solver in the 3D CFD model and keeping theboundary temperatures constant during these intervals, an approximation is made. This approach considers theheat dissipation process to occur through a series of ”small steady state processes” whereby the temperaturesof the engine solids stay constant. The shorter the interval will be for data exchange between the two software,the more transient will be the nature of the simulation will be in nature. Different exchange rates varying fromonce every 5 seconds to once every 60 seconds are compared for the developed tool in the ”Results” section.

3.4.3 Coupling Strategy

One of the biggest advantages of simulation models over physical testing methods is their ability to analyze amyriad of scenarios while spending minimal physical resources for the changes made. But to analyze thesescenarios with the help of simulations and for any model to be ultimately useful for the industry, reduction incomputational time is highly desired as long as the results are accurate enough. In order to achieve this, anefficient strategy is desired which can eliminate redundancies and wastage of computational resources.

As described earlier, the Java macro reads the parameters in the workspace of the 3D model, more specifically,from the temperature reports set-up in the STAR-CCM+ model and also assigns the values of the temperaturesto the boundaries ?? defined in it. It is upgraded for implementing an active coupling-decoupling algorithmwhich saves computational resources. This is done by splitting the drive cycle into two main phases on the basisof the engine state, i.e when the engine is on and off respectively, and using different simulation routines forthem. Further elaboration of the same is done below by considering the described phases as two separate cases:

• Case 1, Engine is switched on: For the periods when the engine is running, forced convection occursin the engine bay due to the incoming air from cooling inlets and the working of the cooling fan. Theheat lost due to natural convection is insignificant in such a scenario and performing simulations with the3D model would hence be futile. Therefore, the macro detects the engine state from the data exported byGT-SUITE and skips the simulations with the 3D CFD model. Instead, nominal values ranging between500 to 1000 watts are imposed are imposed for the rate of heat loss from the engine solids. Thus, thecomputationally expensive 3D model is effectively paused, resulting in reduced demand for computationalresources.

31

• Case 2, Engine is switched off: This case is refers to the scenarios when the car is standing in aquiescent environment and the engine has been switched off. Buoyancy driven convection and radiationmodes of heat transfer dominate under such conditions and simultaneous coupling between the twomodels is kept active. A convergence criteria is set for the 3D simulations to ensure that they stabilizebefore the values for the heat transfer rate are exported to GT-SUITE. A range of +/-2 Watts for thevariation in values of heat transfer rate between successive iterations of the 3D model is assigned for thispurpose. Once the fluctuation of these values for all the engine solids are found to fall within this range,the simulation is deemed to have converged and the values are exported. To minimize the possibility ofnumerical errors, a minimum of three iterations are enforced and the convergence checks are performedonly after the third iteration for any instance of simulation with the 3D CFD model.

This strategy is better illustrated with the help of the schematic depicted in figure 3.20 while the completecode for the Java macro is provided in the appendix A.1

Figure 3.20: Schematic of the coupling strategy employed by the Java macro for the exchange of data betweenthe two software

32

3.4.4 Customized Drive Cycle

Finally, for the analysis of the coupled simulation tool, a customized drive cycle is constructed. This cycle,illustrated in figure 3.21, utilizes the Class 3 Worldwide harmonized Light vehicles Test Cycle (WLTC), whichis presently the standard for passenger vehicles with a power to weight ratio greater than 34 kW/Tonne, s inthe case of our subject vehicle. It consists of an initial driving phase followed by a long shut-down phases ofapproximately 2 hours. The cycle ends with another similar sequence albeit with shorter driving and shut-downphases. It has an overall duration of 5 hours. The primary motive behind the construction of this cycle is toanalyze the stability and robustness of the coupling tool during both driving and long shut-down phases, aswell as the instances of transition between them. The vehicle is subjected to this cycle and the temperature ofthe oil in the engine sump is monitored over time.

Figure 3.21: Customized drive cycle constructed for analysis of the coupled simulation tool

3.4.5 Ambient Conditions

For the simulations performed with the coupled model, a set of ambient conditions are employed as summarizedin table 3.4 below:

Parameter ValueOutside Ambient Temperature (Car Running) 20◦C

Garage Ambient Temperature (Car Parked) 40◦C

Ambient Pressure Atmospheric Pressure(For all cases)

Table 3.2: Ambient conditions employed for coupled simulations

33

4

Results and Discussion

4.1 Predicted Oil Temperatures

The final coupled model is able to predict the variation of the mass averaged temperature of oil in engine sumpover the duration of the defined custom drive cycle 3.21. A plot depicting this variation, when an exchangeinterval of 15 seconds was used for the coupled model, is presented in figure 4.1.The oil temperatures can beobserved to fall after the two instances of engine shut-down occurring after 30 and 160 minutes respectivelyafter the start of the cycle. This fall is expectedly sharp immediately after the shut-down and reduces at alower rate later on. However the rate of fall in temperature is almost constant after a point while the commonintuition suggests that it should be rather digressive in nature. This is primarily due to the scale of variation inthe temperature values which is not so large, and hence, there is a lesser fall in the rate of temperature loss.Another possible explanation for this is the steady state approximation, that had to be employed with the3D model. This approximation requires the heat transfer rates to be constant for a series of intervals of timeinstead of varying continuously with time. Also, the convergence criteria used for the 3D CFD model, wherebythe the heat transfer rates are allowed to vary in the range +/- 2 Watts, could have allowed for errors in theresults and stricter tolerances for the same can be explored.

Figure 4.1: Variation of temperature of oil in the sump for simulation number discussed in section

4.2 Sensitivity Analysis

In order to study the effect of approximation made with respect to the transient nature of heat dissipation inthe engine bay, a sensitivity analysis is performed. This is done to gauge the extent to which the steady-stateapproximation can be implemented without significant deviation in the results. Five simulations were runwhereby intervals of 5, 15, 30, 45 and 60 seconds of physical time were employed for data exchange between thetwo software. All simulations were performed with custom 5-hour drive cycle constructed before.The variationof the mass-averaged oil sump temperatures from this study are presented in figure 4.2.

The rate at which heat dissipates from the external surface of the engine solids is a function of the effectiveheat transfer coefficient and the temperature difference applicable for their interface with the surrounding air.

34

Figure 4.2: Mass-averaged oil sump temperatures from coupled simulations

As their temperature goes down after the engine shut-down, the heat loss rate also reduces. Hence, from atheoretical perspective, it is expected that different exchange intervals between the two software will lead todifferent temperatures of the oil. For the simulations where data is exchanged after longer intervals, such as60 seconds, the rate of heat is loss to the surroundings that is imposed on the engine solids is maintainedat a higher value for longer intervals than in the actual scenario. In other words, the simulations performedwith higher rate of data exchange would are ”more transient” in nature and closer to what happens in reality.The simulation with exchange interval of 5 seconds can hence be considered as the benchmark among thesesimulations from the point of view of accuracy.

This is evident in the plotted results 4.2. At the same time, it can also be noticed that the difference in thepredicted temperature values diminish after a point. The variation in oil temperature values for the simulationsperformed with 5, 10 and 15 seconds of exchange interval are found to be less than 1◦ C for the entire cycle.On the other hand, the simulations with 60 seconds of exchange interval showed upto 7 degrees of differencefrom the values predicted by the simulation with exchange interval of 5 seconds. This leads to the conclusionthat performing coupled simulations with the developed tool with data exchange intervals of 30 seconds or lessare sufficiently precise for most applications.

35

Figure 4.3: Mass-averaged exhaust manifold temperatures from coupled simulations

4.3 Thermal Soak Phenomenon

Thermal soak is a phenomenon that results from prolonged driving of a vehicle at high load, followed byshutting off the engine. It results in increased temperatures of the engine solids and surrounding components,namely the parts in close proximity to the exhaust manifold, after the engine is shut down. This occurs becausethe exhaust manifold is still very hot when the engine is shut down hence the parts close to it continue toreceive energy in form of radiation. Also, once the engine is shut down, the cooling system ceases its dutyof cooling down the engine. The rate at which heat is lost from these engine solids through convection isovershadowed by the rate at which they gain the heat energy being radiated by the hot exhaust manifold.

As the 3D CFD model of the developed simulation takes into account the effects of the convection andradiation, the phenomenon of thermal soak can be observed in the simulation results. The variation of massaveraged temperature of the engine cylinder head are presented in figure 4.4. The rise in temperature of thecylinder head during the 10 minutes (approximately) that follow the two occurences of engine shut down in thedrive cycle can be observed in this plot.

Figure 4.4: Thermal soak phenomenon observed for the cylinder head

36

5

Concluding RemarksThis work presents the development of a coupled 1D-3D simulation tool for numerical analysis of the radiationand natural convection driven conjugate heat transfer in the engine bay of a passenger vehicle. Following arethe primary specifications and functionality of this tool:

• A fully integrated, 1D model of the powertrain cooling systems of the Volvo S80 sedan in GT-SUITEsoftware is utilized. This model has been validated against experimental data. A mean value model ofthe engine was employed for computing the rejected combustion heat. The exhaust manifold has beenmodelled as a thermal mass that can receive heat from the exhaust gases and dissipate it to its externalsurroundings. A time-dependent profile of the mass averaged temperature of the exhaust manifold andthe oil in the sump are captured by this model.

• A 3D CFD model of the natural convection and radiation modes of heat transfer in the engine bay isemployed in STAR-CCM+.

• Communication of surface temperatures of the engine solids and their heat transfer rates, between thetwo software, is achieved through a Java macro which uses comma separated values files to exchangethese data.

• The coupled simulation tool is versatile and can be used to perform simulations for any user defined drivecycle.

• Runtimes approximately equal to 4 X Physical-time were achieved for the coupled model was whileutilizing 200 computing cores for the 3D CFD simulations. This is valid for an exchange interval of 15seconds and a tolerance of +/- 2 Watts in the heat transfer rates calculated by the 3D model.

• Through further improvements, the developed tool has potential for quantifying the fuel saving benefitsof thermal encapsulations for vehicle engines.

37

6

Future WorkThe present study provides a strong basis for further improvements in the following areas:

• Setting up a more detailed 3D CFD model in terms of spatial resolution of the thermal boundaries. Thisimprovement will result in more accurate predictions of the external flow field around the powertraincomponents, improving the overall fidelity of the model.

• Performing transient 3D CFD simulations instead of currently employed steady-state approximation, andcomparing the results from the coupled simulations for these two setups.

• Collecting experimental data for the surface temperatures of the engine solids for validation purposes andfurther calibration of the models to improve their reliability and robustness.

• Replacing the existing Boolean fan controller in the 1D model with a more comprehensive one to improvethe accuracy of the predicted temperatures for the cooling system.

• Improving the thermostat model to make it possible to control its functioning based on coolant pressure,independent of the cooolant temperatures.

• Calibration of the turbocharger oil circuit branch to better predict the heat rejected by them.

38

Bibliography[1] I. Dincer H.S. Hamut and G.F. Naterer. Performance assessment of thermal management systems for

electric and hybrid electric vehicles. int. j. energy res., 37: 1–12. doi:10.1002/er.1951. Technical report,Internationl Journal of Energy Research, 2013.

[2] Ulf Nilsson Emil Ljungskog. Cfd for underhood modeling: Development of an efficient method, master’sthesis 2014:37, issn 1652-8557. Technical report, Chalmers University of Technology, 2014.

[3] Mercedes ECO thermo engine cover. http://www.greencarcongress.com/2014/03/20140318-eco.html.

[4] Parviz Merati, Charles Davis, K-H Chen, and JP Johnson. Underhood buoyancy driven flow—anexperimental study. Journal of Heat Transfer, 133(8):082502, 2011.

[5] Blago B Minovski, Lennart Lofdahl, and Peter Gullberg. Numerical investigation of natural convectionin a simplified engine bay,” sae technical paper 2016-01-1683, 2016, doi:10.4271/2016-01-1683. Technicalreport, SAE Technical Paper, 2016.

[6] Yunus A Cengel, Sanford Klein, and William Beckman. Heat Transfer: A practical approach, volume 2ndEdition. McGraw-Hill New York, 1998.

[7] Mahajan Sammakia Gebhart, Jaluria. Buoyancy Induced Flows and Transport, volume Textbook Edition.Taylor and Francis, 2003.

[8] Water Pump. http://asc-ind.com/wp-content/uploads/2010/10/PUMP-w-raw-mat-callouts.jpg.

[9] Thermostat. http://www.ustudy.in/sites/default/files/images/thermostat.jpg.

[10] Dinakar P. Heilbronn D. Jurik J. Kumar S. Rajeeve G. Sarovec M. Coupled unsteady simulations ofpowertrain cooling systems. Technical report, Chalmers, 2015.

[11] Intercooler. http://www.b15u.com/attachment.php?attachmentid=57&amp;stc=1&amp;d=

1248287952.

[12] Radiator. http://radiatorx.blogspot.se/2013/01/car-radiator-fluid.html.

[13] Minovski, Blago B and Lofdahl, Lennart: Gothenburg, Chalmers University of Technology, 2013. Diplomawork: Study of software system integration for transient simulation of future cooling system for heavytruck applications , ISSN 1652-8557; 2013:40, 2013. SAE Technical Paper, 2014 .

[14] Engine Cooling Fan. http://repairpal.com/images/managed/content_images/encyclopedia/CM_

Engine/Engine_Cooling_Fan_09.11.png.

[15] Intercooler. http://www.zoombd24.com/wp-content/uploads/2015/02/Expansion-tank.png.

[16] Gamma Technologies official website. http://www.gtisoft.com.

[17] Mattias Wahde. Biologically inspired optimization methods: An introduction. WIT press, 2008.

[18] Morel, T and others, Manual for GT Power version 7.5, Gamma Technologies, 2014.

[19] Ronald L Iman. Latin hypercube sampling. Encyclopedia of quantitative risk analysis and assessment,2008.

I

A

Appendix

A.1 Java Macro Code

package WriteTest;

import java.util.*;

import java.io.*;

import star.base.neo.*;

import star.base.report.Report;

import star.common.*;

import star.vis.*;

import star.flow.*;

import star.energy.*;

public class CyberdyneS103 extends StarMacro {int k=0;

public void execute() {

execute0();

}

private void execute0() {

Simulation simulation_0=getActiveSimulation();

File f = new File(resolvePath("Output.csv"));

File fi = new File(resolvePath("Input.csv"));

double doubleValue1=0.0; // Declaring and initializing engine state variable

double doubleValue2=0.0;

double doubleValue3=0.0;

double doubleValue4=0.0;

double doubleValue5=0.0;

double doubleValue6=0.0;

double doubleValue7=0.0;

double doubleValue8=0.0;

for (int i = 1; i <= 1600; i++) { // This loop specifies the maximum allowed number of

exchanges between GT-SUITE and StarCCM

if (i>1) { //The Input file written by GT-SUITE is read before the second

exchange in order to allow StarCCM to make at least one iteration before being

dependent on GT-SUITE

if (!fi.exists()) { // Checking if Input.csv exists

simulation_0.println("Waiting for GT-SUITE to write Input.csv at exchange number " +

(i-1));

}

while (!fi.exists()) { // Wait until GT-SUITE writes Input.csv

try {

Thread.sleep(1000);

} catch (InterruptedException e) {

//Handle exception

}

}

II

try { //The try-catch statement ensures that the code doesnt crash in

case something goes wrong

FileReader fr = new FileReader(fi);

BufferedReader br = new BufferedReader(fr);

Scanner sc = new Scanner(br).useLocale(Locale.ENGLISH);

sc.useDelimiter(","); //Defining delimiter for .csv format

int j=1; //Counter for row number

while (sc.hasNextLine() ) {

if (j==2) {

doubleValue2 = sc.nextDouble(); // Reading the first double on the second

row of the .csv file

simulation_0.println("Engine head temperature read in Kelvin " +

doubleValue2);

// Updating boundary conditions in STARCCM simulation

Region region_0 = simulation_0.getRegionManager().getRegion("Region");

Boundary boundary_0 =

region_0.getBoundaryManager().getBoundary("Engine_head");

StaticTemperatureProfile staticTemperatureProfile_0 =

boundary_0.getValues().get(StaticTemperatureProfile.class);

Units units_0 = ((Units) simulation_0.getUnitsManager().getObject("K"));

staticTemperatureProfile_0.getMethod(ConstantScalarProfileMethod.class).getQuantity().setUnits(units_0);

staticTemperatureProfile_0.getMethod(ConstantScalarProfileMethod.class).getQuantity().setValue(doubleValue2);

//

String s;

s = sc.next();

doubleValue3 = Double.parseDouble(s);

simulation_0.println("Engine block temperature read in Kelvin " +

doubleValue3);

//Updating boundary coinditions in STARCCM simulation

Boundary boundary_1 =

region_0.getBoundaryManager().getBoundary("Engine_block");

StaticTemperatureProfile staticTemperatureProfile_1 =

boundary_1.getValues().get(StaticTemperatureProfile.class);

Units units_1 = ((Units) simulation_0.getUnitsManager().getObject("K"));

staticTemperatureProfile_1.getMethod(ConstantScalarProfileMethod.class).getQuantity().setUnits(units_1);

staticTemperatureProfile_1.getMethod(ConstantScalarProfileMethod.class).getQuantity().setValue(doubleValue3);

//

String b;

b = sc.next();

doubleValue4 = Double.parseDouble(b);

simulation_0.println("Oil sump temperature read in Kelvin " +

doubleValue4);

//Updating boundary coinditions in STARCCM simulation

Boundary boundary_2 =

region_0.getBoundaryManager().getBoundary("Oil_sump");

StaticTemperatureProfile staticTemperatureProfile_2 =

boundary_2.getValues().get(StaticTemperatureProfile.class);

Units units_2 = ((Units) simulation_0.getUnitsManager().getObject("K"));

staticTemperatureProfile_2.getMethod(ConstantScalarProfileMethod.class).getQuantity().setUnits(units_2);

staticTemperatureProfile_2.getMethod(ConstantScalarProfileMethod.class).getQuantity().setValue(doubleValue4);

String d;

d = sc.next();

doubleValue1 = Double.parseDouble(d);

simulation_0.println("Current Engine State is " + doubleValue1);

String bb;

III

bb = sc.next();

doubleValue5 = Double.parseDouble(bb);

simulation_0.println("Manifold temperature read in Kelvin " +

doubleValue5);

//Updating boundary coinditions in STARCCM simulation

Boundary boundary_3 =

region_0.getBoundaryManager().getBoundary("Exhaust_manifold");

StaticTemperatureProfile staticTemperatureProfile_3 =

boundary_3.getValues().get(StaticTemperatureProfile.class);

Units units_3 = ((Units) simulation_0.getUnitsManager().getObject("K"));

staticTemperatureProfile_3.getMethod(ConstantScalarProfileMethod.class).getQuantity().setUnits(units_3);

staticTemperatureProfile_3.getMethod(ConstantScalarProfileMethod.class).getQuantity().setValue(doubleValue5);

String ot;

ot = sc.next();

doubleValue6 = Double.parseDouble(ot);

simulation_0.println("Oil temperature read in Kelvin " + doubleValue6);

String ch;

ch = sc.next();

doubleValue7 = Double.parseDouble(ch);

simulation_0.println("Cylinder head temperature read in Kelvin " +

doubleValue7);

String tm;

tm = sc.next();

doubleValue8 = Double.parseDouble(tm);

simulation_0.println("Time read in s " + doubleValue8);

if (doubleValue1==1){

simulation_0.println("The engine is currently running, feeding standard

values for heat transfer...");

k=1;

}

else { simulation_0.println("The engine is currently shut down, STAR-CCM+

will compute the HTR values...");

k=2;}

}

// Deleting Input.csv after reading it

else {

sc.nextLine();

}

j++;

}

br.close();

fr.close();

}

catch (Exception ex) {

simulation_0.println("ERROR reading file."); // In case anything goes wrong

}

// Deleting Input.csv after reading it

try{

if(fi.delete()){

IV

simulation_0.println(fi.getName() + " is deleted!");

}else{

simulation_0.println("Delete operation is failed.");

}

}catch(Exception e){

e.printStackTrace();

}

}

// Checking if Output.csv exists

if (f.exists()) {

simulation_0.println("Waiting for GT-SUITE to read and delete Output.csv at exchange

number " + (i-1));

}

while (f.exists()) { // Wait until GT-SUITE deletes Output

try {

Thread.sleep(1000);

} catch (InterruptedException e) {

//Handle exception

}

}

if(k==2){

simulation_0.println("Preparing output for GT-SUITE for exchange number " +(i-1));

double residual_os=0.0;

double residual_eb=0.0;

double residual_em=0.0;

simulation_0.getSimulationIterator().step(2,true);

Collection<Report> colrep = simulation_0.getReportManager().getObjects();

double[] repValchecker = new double[colrep.size()];

simulation_0.println("\n\nSimulation will now iterate till the HTR values

stabilize...");

for (int z = 1; z <= 60; z++){

simulation_0.println("Starting iteration number " + z + " for exchange process number

" +(i-1)+’\n’);

for (Report rep : colrep){

if (rep.getPresentationName().contains("3")) {

repValchecker[0] = rep.getReportMonitorValue();}}

for (Report rep : colrep){

if (rep.getPresentationName().contains("2")) {

repValchecker[2] = rep.getReportMonitorValue();}}

for (Report rep : colrep){

if (rep.getPresentationName().contains("7")) {

repValchecker[4] = rep.getReportMonitorValue();}}

simulation_0.getSimulationIterator().step(1,true);

V

for (Report rep : colrep){

if (rep.getPresentationName().contains("3")) {

repValchecker[1] = rep.getReportMonitorValue();}}

for (Report rep : colrep){

if (rep.getPresentationName().contains("2")) {

repValchecker[3] = rep.getReportMonitorValue();}}

for (Report rep : colrep){

if (rep.getPresentationName().contains("7")) {

repValchecker[5] = rep.getReportMonitorValue();}}

simulation_0.println("\nIteration values saved...");

residual_os=(repValchecker[1]-repValchecker[0]);

residual_eb=(repValchecker[3]-repValchecker[2]);

residual_em=(repValchecker[5]-repValchecker[4]);

simulation_0.println("Current value of Oil Sump HTR residual is " +residual_os);

simulation_0.println("\nCurrent value of Engine Block HTR residual is " +residual_eb);

simulation_0.println("\nCurrent value of Exhaust manifold HTR residual is "

+residual_em);

if (residual_os<2 && residual_os>-2 && residual_eb<2 && residual_eb>-2 &&

residual_em<2 && residual_em>-2)

{simulation_0.println("Residuals annulled! Breaking iteration loop now...");break;}

}

}

if(i>1){writeToFile(simulation_0); writeToFile1(simulation_0, doubleValue5, doubleValue6,

doubleValue7, doubleValue8);}

}

}

private void writeToFile(Simulation simulation_0) {

Collection<Report> colrep = simulation_0.getReportManager().getObjects();

double[] repVal = new double[colrep.size()];

String line = "";

File buffer = new File(resolvePath("Buffer.csv"));

File forig = new File(resolvePath("Output.csv"));

VI

try {

BufferedWriter bw = new BufferedWriter (new FileWriter(buffer,true));

String headerline;

headerline =

"\"HeatTransfer1\",\"HeatTransfer2\",\"HeatTransfer3\",\"HeatTransfer7\",\"HeatTransfer5\",\"HeatTransfer6\"";

bw.write(headerline);

bw.newLine();

if (k==1){

for (Report rep : colrep) {

if (rep.getPresentationName().contains("1")) {

repVal[0] = -1200;

repVal[4] = -800;

}

if (rep.getPresentationName().contains("2")) {

repVal[1] = -1500;

}

if (rep.getPresentationName().contains("3")) {

repVal[2] = -600;

}

if (rep.getPresentationName().contains("7")) {

repVal[3] = -5000;

}

}

}

else { for (Report rep : colrep) {

if (rep.getPresentationName().contains("1")) {

repVal[0] = Math.round(rep.getReportMonitorValue()*200/3)/100;

repVal[4] = Math.round(rep.getReportMonitorValue()*100/3)/100;

}

if (rep.getPresentationName().contains("2")) {

repVal[1] = Math.round(rep.getReportMonitorValue()*100/2)/100;

}

if (rep.getPresentationName().contains("3")) {

repVal[2] = Math.round(rep.getReportMonitorValue()*100)/100;

}

if (rep.getPresentationName().contains("7")) {

repVal[3] = Math.round(rep.getReportMonitorValue()*100)/100;

}

}

}

simulation_0.println("Outputs to GT-SUITE:" + repVal[0] + ", " + repVal[1] + ", " +

repVal[2] + ", " + repVal[3] + ", " + repVal[4] +’\n’ + ’\n’ + ’\n’ +’\n’);

VII

line = String.valueOf(repVal[0]) + "," + String.valueOf(repVal[1]) + "," +

String.valueOf(repVal[2]) + "," + String.valueOf(repVal[3])+ "," +

String.valueOf(repVal[4]);

bw.write(line);

bw.newLine();

bw.close();

}catch (IOException iOException) {

}

buffer.renameTo(forig);

}

//

private void writeToFile1(Simulation simulation_0, Double doubleValue5, Double doubleValue6,

Double doubleValue7, Double doubleValue8) {

String line = "";

try {

File f = new File(resolvePath("OutputTemp.csv"));

BufferedWriter bw = new BufferedWriter (new FileWriter(f,true));

String headerline;

headerline = "\"Time\",\"Oil temp\",\"Exhaust temp\",\"Head temp\"";

bw.write(headerline);

bw.newLine();

simulation_0.println("Logged Data:" + doubleValue8 + ", " + doubleValue6 + ", " +

doubleValue5 + ", " + doubleValue7 + ’\n’ + ’\n’ + ’\n’ +’\n’);

line = "" + doubleValue8 + "," + doubleValue6 + "," + doubleValue5 + "," + doubleValue7;

bw.write(line);

bw.newLine();

bw.close();

}catch (IOException iOException) {

}

}

}

VIII

A.2 Distribution of Content

The contribution of the students writing this report to the content presented in it is shown in table A.1 belowfor future references:

Parameter AuthorFront Matter (Cover, Abstract, Preface & Introduction) Saket KumarNomenclature Daniel Pascual and Saket KumarTheoretical and Modelling Background (Pages: 2, 5-10) Daniel PascualTheoretical and Modelling Background (Pages: 3-4, 11) Saket KumarMethods (Pages 18-26) Daniel PascualMethods (Pages 12-17, 27-31) Saket KumarResults and Discussion (Pages 32-34) Saket KumarConcluding Remarks (Page 35) Saket KumarFuture Work (Page 36) Saket KumarAppendix (Page I-VIII) Saket KumarBibliography Daniel Pascual and Saket Kumar

Table A.1: Distribution of content in the report

IX